In this application, we propose to address PQ10: As we improve methods to identify epigenetic changes that occur during tumor development, can we develop approaches to discriminate between driver and passenger epigenetic events?. Epigenetic mechanisms exert a strong effect on gene expression potential, and have been shown to undergo widespread change in most human cancers. In contrast to most mutational events, epigenetic events often display a high degree of correlation, with a large number of defined alterations that appear to be passenger events without functional contribution to the cancer process. We propose to develop an integrated computational and experimental validation pipeline to identify epigenetic driver events in cancer. In Aim 1 we will develop a probabilistic framework for predicting and prioritizing candidate epigenetic driver loci. This approach is unique in that it fully integrates the wealth of available data, using complementary data types derived from primary genomic data, experimental data, and supporting curated information, resulting in a composite Epigenetic Driver Score (EDS), reflecting the posterior probability that each gene is an epigenetic driver. Aim 2 will provide experimental data on epigenetic addiction, using cell lines depleted of DNA methyltransferases, and thus selected to retain only the most essential silencing events, in addition to data obtained with embryonic and adult stem-cell and progenitors. These experimental data sets will be used to complement primary epigenomic data we have generated in the context of TCGA, to provide Epigenetic Driver Scores for each locus in each tumor type, using the methodology developed in Aim 1. In Aim 3a we will functionally test the top-ranked candidate epigenetic drivers of colon, breast, and lung cancer in vitro, by reintroducing expression of candidate genes into appropriate human cancer cells lines containing the relevant silencing events. These experiments will be complemented by shRNA approaches in cell lines to modulate the functional expression of the candidate epigenetic drivers. In vitro proliferation and apoptosis assays will be used to assess phenotypic effects. In Aim 3b we will assess the functional contributions of the candidate epigenetic drivers in vivo, using the stable cell lines created in Aim 3a in xenograft mouse models. The results of these validation experiments will be used to iteratively train the EDS model. Given the sensitivity of learning algorithms to their training data, we anticipate an improvement in performance as the number of training examples increases. By performing data-driven modeling in a probabilistic framework and computationally- directed experimentation the available data will be utilized to the fullest extent, while allowing for the addition of new data types and expert curation. The role of epigenetic events in cancer is increasingly appreciated, but the challenge of distinguishing drivers from passengers has not yet been adequately addressed. The systematic validation pipeline proposed here will address a large unmet need, and yield insights into the complementary roles of epigenetic and genetic events in key signaling pathways that drive tumorigenesis. PUBLIC HEALTH RELEVANCE: Cancer arises from loss of the growth control of cells. Although most attention has focused on genetic mutations as the origin of these defects, in recent years, it has become increasingly appreciated that changes in the epigenetic control of gene activity also contribute to this loss of growth control. This application addresses an unmet need to develop methods of finding out which epigenetic changes contribute directly to cancer formation.